How Three Universities Developed Their Chatbots

According to a recent EDUCAUSE report, 37% of colleges and universities provide institutionwide licenses for chatbots, and 14% have their own homegrown bots.

Implementing these tools requires IT and educational leaders to carefully consider factors such as data privacy, pedagogy and infrastructure, while the technology itself continues to evolve at a dizzying pace.

Here’s how three institutions are navigating the challenge.

University of Michigan Offers Several AI Services to Staff and Students

In August 2023, before OpenAI had even introduced its enterprise offerings, the University of Michigan launched UM-GPT, becoming the first higher education institution to offer AI services at scale.

The creation of the platform, which provides secure access to commercially available large language models through the university’s private Microsoft Azure environment, illustrates Michigan’s proactive approach to embracing AI while maintaining control over the university’s data.

“There are going to be both opportunities and challenges,” says Ravi Pendse, vice president for IT and CIO. “But on balance, this will be a force for positive disruption.”

In the past two years, the university has also launched U-M Maizey (a codeless platform that allows users to create custom AI experiences by uploading their own data sets) and Go Blue (a mobile app that serves as an AI assistant for campus life). Right now, Pendse’s team is getting ready to launch GoToCollege, a public-facing chatbot that will connect students across the country with scholarship opportunities.

“Now, students have access to tutoring 24/7,” Pendse says. “They can find protein-rich meals on campus or up-to-date information about bus routes. All that information is coming together through AI assistants.”

The tools have seen rapid adoption across campus. Faculty can integrate Maizey with the university’s learning management system to set up course-specific AI tutoring in just a few minutes. There are now more than 3,500 Maizey instances in production across academic departments and in operational offices such as procurement and HR. The platform uses retrieval-augmented generation to prevent hallucination.

To avoid vendor lock-in, Michigan runs its AI workloads in a variety of environments, including on-premises infrastructure. When users interact with UM-GPT, they can select which LLM they want to use via a dropdown menu. The university’s approach not only keeps student and faculty data out of public LLM training models but also ensures equitable access for those who might not be able to afford monthly AI subscription fees. Additionally, Michigan’s platforms support screen readers, providing access for people who are visually impaired.

Pendse thinks all universities should require students to take an AI course before they graduate or at least offer their students significant exposure to AI training and tools. “They’re going out into a world where they’re competing with people who know how to use these tools,” he says. “Generative AI is going to be the most impactful technology of this century.”

University of California, Irvine Allows Users to Create Their Own Bots

When ChatGPT 3.5 took the world by storm, leaders at the University of California, Irvine asked the IT department to form a working group to investigate the technology.

That group spent two months over the summer coming up with recommendations, mainly about how the IT department itself could use AI to enhance efficiency in its own operations. But as the technology grew, so did the working group’s mandate.

“We were asked to deploy something across campus,” recalls Sarkis Daglian, director of AI, cloud and client solutions at UC Irvine. “We were able to go from nothing to a standing service within three months.”

READ MORE: Assess your institution’s AI readiness before embarking on a new project.

That urgency was driven by several factors. For one, university leaders wanted to create a level playing field for all students, rather than leaving AI services to those who could afford monthly subscriptions. And they wanted to prepare students for a technology that is already transforming the workplaces where they will soon be seeking jobs. Also, Daglian says, there was a simple recognition that AI was going to be a transformative technology on par with previous trends like the rise of the PC and the internet.

“We wanted to position ourselves on the cutting edge,” Daglian says.

Chris Price, AI architect and lead AI developer at UC Irvine, notes that researchers can more freely experiment with AI when they have access to a secure, private platform that doesn’t share user data. “Before, many stakeholders were afraid to even step into the water,” he says. “Those are the exact people we need to use this technology to help shape the future.”

In January 2024, UC Irvine launched ZotGPT, a free AI tool for students, faculty and staff built using resources from Azure AI, Amazon Web Services and open web technologies. The platform offers access to multiple large language models, as well as API access for researchers, and it has attracted 15,000 unique users. The university is testing a handful of additional AI tools, including ZotGPT Creator, which will allow users to create their own customized AI chatbots. Another tool, ZotGPT ClassChat, launched in January and allows instructors to incorporate course materials with LLMs for students at much higher detail than a standard chatbot.

The technology continues to move quickly, but Price says that’s not a reason to sit back and wait for solutions to shake out. “In fact, it’s quite the opposite,” he says. “That requires us to work even harder and faster, so we can prepare our community for the next big disruption.”

Arizona State University Bots Support Teaching and Learning

“Guidelines foster innovation,” says Elizabeth Reilley, executive director of AI acceleration for ASU Enterprise Technology, the technology and IT division of Arizona State University. “When people don’t have guidance on what’s safe, they’re often afraid to do anything.”

LEARN MORE: Are your chatbots giving away more information than they should?

This philosophy has led ASU to proactively roll out an expansive — and constantly growing — set of AI solutions. The university’s CreateAI Platform, launched in October 2023, gives students, faculty and staff secure access to more than 50 generative AI models. The platform offers a tool called MyAI Builder that lets users create their own AI projects by adding specific system prompts and instructions to LLMs.

ASU’s AI environment is largely built on AWS infrastructure, with integrations between the environment and major generative AI models.

The university also supports several purpose-built chatbots, built on ChatGPT Edu, which was offered through the university’s AI Innovation Challenge. One of these, named Sam, lets students in behavioral health practice their patient interviewing skills. Another lets students debate with famous philosophers from history.

UP NEXT: Miami Dade College improves business operations with AI.

“We take a comprehensive approach to supporting AI across teaching and learning, across research and across how people work every day,” says ASU Deputy CIO Kyle Bowen. “We’ve been able to put enterprise-class AI tools into the hands of the faculty, students and staff who are shaping what the future of AI is going to look like.”

ASU has invested heavily in promoting AI literacy, with 3,000 faculty members participating in professional development sessions. Also, any faculty or staff member can sign up for the university’s CreateAI Lab, which includes online community groups and weekly Q&A sessions.

“It's a learning environment, not just from our team to others, but also a peer learning environment,” Reilley says. “So, if folks are struggling, they can share their struggles. If somebody found something that was really interesting that took their work to the next level, they can share that as well.”

“The future of AI is still unfolding,” Bowen says. “To be at the forefront, people need hands-on access. New AI pedagogies and research methods are being invented right now in classrooms and labs around the world, and we need to create the space to innovate and engage in a hands-on way.”